Gon4l regulates notochord boundary formation and cell ...€¦ · notochord boundary via...

ARTICLE

Gon4l regulates notochord boundary formation andcell polarity underlying axis extension by repressingadhesion genesMargot L.K. Williams1, Atsushi Sawada1,2, Terin Budine1, Chunyue Yin2,3, Paul Gontarz1

& Lilianna Solnica-Krezel1,2

Anteroposterior (AP) axis extension during gastrulation requires embryonic patterning and

morphogenesis to be spatiotemporally coordinated, but the underlying genetic mechanisms

remain poorly understood. Here we define a role for the conserved chromatin factor Gon4l,

encoded by ugly duckling (udu), in coordinating tissue patterning and axis extension during

zebrafish gastrulation through direct positive and negative regulation of gene expression.

Although identified as a recessive enhancer of impaired axis extension in planar cell polarity

(PCP) mutants, udu functions in a genetically independent, partially overlapping fashion with

PCP signaling to regulate mediolateral cell polarity underlying axis extension in part by

promoting notochord boundary formation. Gon4l limits expression of the cell–cell and

cell–matrix adhesion molecules EpCAM and Integrinα3b, excesses of which perturb the

notochord boundary via tension-dependent and -independent mechanisms, respectively. By

promoting formation of this AP-aligned boundary and associated cell polarity, Gon4l coop-

erates with PCP signaling to coordinate morphogenesis along the AP embryonic axis.

DOI: 10.1038/s41467-018-03715-w OPEN

1 Department of Developmental Biology, Washington University School of Medicine, Saint Louis, MO 63110, USA. 2Department of Biological Sciences,Vanderbilt University, Nashville, TN 37235, USA. 3Present address: Division of Pediatric Gastroenterology, Hepatology, and Nutrition, Cincinnati Children’sHospital, Cincinnati, OH 45229, USA. Correspondence and requests for materials should be addressed to L.S-K. (email: [email protected])

NATURE COMMUNICATIONS | (2018) 9:1319 | DOI: 10.1038/s41467-018-03715-w |www.nature.com/naturecommunications 1

1234

5678

90():,;

mailto:[email protected]

www.nature.com/naturecommunications


Gastrulation is a critical period of animal developmentduring which the three primordial germ layers—ecto-derm, mesoderm, and endoderm—are specified,

patterned, and shaped into a rudimentary body plan. Duringvertebrate gastrulation, an elongated anteroposterior (AP) axisemerges as the result of convergence and extension (C&E), aconserved set of gastrulation movements characterized by theconcomitant AP elongation and mediolateral (ML) narrowing ofthe germ layers1,2. C&E is accomplished by a combination ofpolarized cell behaviors, including directed migration and MLintercalation behavior (MIB)3–5. During MIB, cells elongateand align their bodies and protrusions in the ML dimensionand intercalate preferentially between their anterior and posteriorneighbors5. This polarization of cell behaviors with respect tothe AP axis is regulated by planar cell polarity (PCP) and othersignaling pathways6–10. Because these pathways are essentialfor MIB and C&E but do not affect cell fates7,10,11, othermechanisms must spatiotemporally coordinate morphogenesiswith embryonic patterning to ensure normal development. BMP,for example, coordinates dorsal–ventral axis patterning withmorphogenetic movements by limiting expression of PCPsignaling components and C&E to the embryo’s dorsal side12.In general, though, molecular mechanisms that coordinategastrulation cell behaviors with axial patterning are poorlyunderstood, and remain one of the key outstanding questions indevelopmental biology.

Epigenetic regulators offer a potential mechanism by whichbroad networks of embryonic patterning and morphogenesisgenes can be co-regulated. Epigenetic modifiers form proteincomplexes with chromatin factors that are thought to regulatetheir binding at specific genomic regions in context-specificways13. The identities, functions, and specificities of chromatinfactors with roles during embryogenesis are only now beingelucidated, and some have described roles in cell fate specificationand embryonic patterning14. One such chromatin factor isGon4l, whose homologs have conserved roles in cell cycleregulation and/or embryonic patterning in plants, worms, flies,mice, and fish15–19. However, the contribution of Gon4l or anyother chromatin factor to morphogenesis is particularly poorlyunderstood.

Here, we demonstrate a role for zebrafish Gon4l, encoded byugly duckling (udu), as a regulator of embryonic axis extensionduring gastrulation. udu was identified in a forward geneticscreen for enhancers of short axis phenotypes in PCP mutants,but we find it functions in parallel to PCP signaling. Instead,complete maternal and zygotic udu (MZudu) deficiency producesa distinct set of morphogenetic and cell polarity phenotypes thatimplicate the notochord boundary in ML cell polarity and cellintercalation during C&E. Extension defects in MZudu mutantsare remarkably specific, as internalization, epiboly, andconvergence gastrulation movements occur normally. Geneexpression profiling reveals that Gon4l regulates expression of alarge portion of the zebrafish genome, including genes with rolesin housekeeping, patterning, and morphogenesis. Furthermore,Gon4l-associated genomic loci are identified by DNA adeninemethyltransferase (Dam) identification20,21 paired with high-throughput sequencing (DamID-seq), revealing both positiveand negative regulation of putative direct targets by Gon4l.Mechanistically, we find that increased expression of epcam anditga3b, direct targets of Gon4l-dependent repression duringgastrulation, each contribute to notochord boundary defectsin MZudu mutants via a distinct molecular mechanism. Thisreport thereby defines a critical role for a chromatin factor in theregulation of gastrulation cell behaviors in vertebrate embryos:by ensuring proper formation of the AP-aligned notochordboundary and associated ML cell polarity, Gon4l cooperates with

PCP signaling to coordinate morphogenesis that extends the APembryonic axis.

ResultsGon4l regulates axis extension during zebrafish gastrulation.To identify previously unknown regulators of C&E, we performeda three generation synthetic mutant screen22,23 using zebrafishcarrying a hypomorphic allele of the PCP gene knypek (kny)/glypican 4, knym8188. F0 wild-type (WT) males were mutagenizedwith N-ethyl-N-nitrosourea (ENU) and outcrossed to WTfemales. The resulting F1 fish were outcrossed to knym818/818

males (rescued by kny RNA injection) to generate F2 familieswhose F3 offspring were screened at 12 and 24 h post-fertilization(hpf) for short axis phenotypes (Fig. 1a). Screening nearly 100 F2families yielded eight recessive mutations that enhanced axisextension defects in knym818/818 embryos (Fig. 1b–e). vu68 wasfound to be a L227P kny allele, and vu64 was a Y219* nonsenseallele of the core PCP gene trilobite(tri)/vangl2, mutations inwhich exacerbate kny mutant phenotypes24, demonstratingeffectiveness of our screening strategy. We focused on vu66/vu66mutants, which displayed a pleiotropic phenotype at 24 hpf,including shortened AP axes, reduced tail fins, and heart edema(Fig. 1d, Supplementary Fig.1b).

Employing the simple sequence repeat mapping strategy25, wemapped the vu66 mutation to a small region on Chromosome 16that contains the udu gene (Fig. 1f). udu encodes a conservedchromatin factor homologous to gon-4 in Caenorhabditiselegans16 and the closely related Gon4l in mammals18,19. Thepreviously described udu mutant phenotypes resemble those ofhomozygous vu66 embryos, including a shorter body axis andabnormal tail fins18,26, making udu an excellent candidate for thisPCP enhancer. Sequencing cDNA of the udu coding region from24 hpf vu66/vu66 embryos revealed a T to A transversion atposition 2261 predicted to change 753Y to a premature STOPcodon (Fig. 1g). Furthermore, vu66 failed to complement aknown udusq1 allele18 (Supplementary Fig. 1c). Together, thesedata establish vu66 as an udu allele and identify it as a recessiveenhancer of axis extension defects in kny PCP mutant gastrulae.

Complete loss of udu function impairs axial extension. Toassess the full role of Gon4l during early development, weeliminated maternal expression of udu18 using germline repla-cement27. The resulting WT females carrying uduvu66/vu66

(udu−/−) germline were crossed to uduvu66/vu66 germline malesto produce 100% embryos lacking both maternal (M) and zygotic(Z) udu function, hereafter referred to as MZudu mutants. Thesemutants appeared outwardly to develop normally until mid-gastrula stages (Fig. 1h, i), but exhibited clear abnormalities at theonset of segmentation (Fig. 1j, k). MZudu−/− gastrulae specifiedthe three germ layers (Fig. 1l–m, Supplementary Fig. 2), formedan embryonic shield marked by gsc expression (SupplementaryFig. 2d), and completed epiboly on schedule (Fig. 1h–i), butsomites were largely absent in mutants (Fig. 1j, k, n, o). AlthoughmyoD expression was observed within adaxial cells by wholemount in situ hybridization (WISH), its expression was notdetected in nascent somites (Fig. 1o), similar to reporteddescriptions of Zudu mutants28. Formation of adaxial cells isconsistent with normal expression of their inducer shh in the axialmesoderm29 (Supplementary Fig. 2h), and ntla/brachyuryexpression in the axial mesoderm was also largely intact (Sup-plementary Fig. 2j). Importantly, MZudu mutants were markedlyshorter than age-matched WT controls throughout segmentation(Fig. 1j, k) and at 24 hpf (Supplementary Fig.3e, h). Althoughincreased cell death was observed in MZudu mutants (as in Zudumutants28), inhibiting apoptosis via injection of RNA encoding

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03715-w

2 NATURE COMMUNICATIONS | (2018) 9:1319 | DOI: 10.1038/s41467-018-03715-w |www.nature.com/naturecommunications


the anti-apoptotic mitochondrial protein Bcl-xL30 did not sup-press their short axis phenotype (Supplementary Fig. 3a-f) asudu-gfp RNA did (Supplementary Fig. 3g-i). These resultsdemonstrate a specific role for Gon4l in axial extension duringgastrulation.

Gon4l regulates formation of the notochord boundary. Time-lapse Nomarski (Fig. 2a, b) and confocal (Fig. 2c, d) microscopyof dorsal mesoderm in MZudu−/− gastrulae revealed reduceddefinition and regularity of the boundary between axial andparaxial mesoderm, hereafter referred to as the notochordboundary, compared to WT (Fig. 2a, b, arrowheads). The ratio ofthe total/net length of notochord boundaries was significantly

higher in MZudu−/− than in WT gastrulae at all-time points(Fig. 2e, two-way ANOVA p < 0.0001), indicative of decreasedstraightness. Interestingly, laminin was detected by immunos-taining at the notochord boundary of both WT and MZudu−/−embryos (Fig. 2f, g), indicating that MZudu mutants form a bonafide boundary, albeit an irregular one. These results demonstratethat Gon4l is necessary for proper formation of the notochordboundary during gastrulation.

Cell polarity and intercalation are reduced in MZudu mutants.In vertebrate gastrulae, C&E is achieved chiefly through MLintercalation of polarized cells that elongate and align theircell bodies with the ML embryonic axis3,5,7,9. To determine

ch16

Phenotype screening at 24 hpf

16/16

14/14

51/51

59/59

WT

MZ

udu

YPC 20S

dlx3/hgg1 myoD

2S 8S

a c

d e

f g

b

n

o

h j l

i k m

ENU

F0

WT

vu66 vu66;knym818

24 hpf

knym818

F1

F2

F3

+/+ +/+

1/2

1/2

z17403

16/843 13/769

0/810

Ugly duckling

NestinMef2dBrevican

CtskOnecutl

Mcl1bSodium channel modifier 1

Cortexin2

Imm

unity

-relat

ed G

TPase

x

x

knym818–/–

knym818+/–;+/+

knym818+/–;VUxl+

VUxl+

1 Mbp

z15431 1CR-1 CR-2 CR-3 PAH-L1 PAH-L2 SANT-L

2055

tu24 sg1vu66457Y

753Y962Y

Stop StopStop

15/81312/749

1.25–1.45 cM(0.75–0.87 Mb)

Fig. 1 A forward genetic screen identifies ugly duckling(udu)/gon4l as a regulator of axis extension in zebrafish embryos. a Schematic of a synthetic screento identify enhancers of the short axis phenotype in knym818/m818 zebrafish mutants. b–e Phenotypes at 24 hpf: wild type (WT) (b), knym818/m818 (c), vu66/vu66 (d), knym818/m818; vu66/vu66 compound mutants (e). Images are representative of phenotypes observed at Mendelian ratios in multiple independentclutches. f Diagram of mapping vu66 mutation to Chromosome 16. Bold numbers below specify the number of recombination events between vu66 andthe indicated loci. g Diagram of the Gon4l protein encoded by the udu locus. Arrowheads indicate residues mutated in vu66 and other described udu alleles.h–k Live WT (h, j) and maternal zygotic (MZ)udu (i, k) embryos at yolk plug closure (YPC) (h–i) and 20 somite stage (j–k). 100 percent of MZudumutantsfrom more than 15 germline-replaced females exhibited the pictured phenotypes. l–m Whole mount in situ hybridization (WISH) for dlx3 (purple) and hgg1(red) in WT (l) and MZudu−/− (m) embryos at two-somite stage. n–o WISH for myoD in WT (n) and MZudu−/− (o) embryos at eight-somite stage.Anterior is to the left in b–e, j–k; anterior is up in h–i, l–o. Fractions indicate the number of embryos with the pictured phenotype over the number ofembryos examined. Scale bar is 500 μm in b–e and 300 μm in h–o

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03715-w ARTICLE




whether cell polarity defects underlie reduced axis extension inMZudu−/− gastrulae, we measured cell orientation (the angle ofa cell’s long axis with respect to the embryo’s ML axis) and cellbody elongation (length-to-width or aspect ratio (AR)) inconfocal time-lapse series of fluorescent membrane-labeled WT(Fig. 3a–d) and MZudu−/− (Fig. 3e–h) gastrulae. Given theirregular notochord boundaries observed in MZudu−/− gas-trulae (Fig. 2d, e), we examined the time course of cell polariza-tion according to a cell’s position with respect to the boundary,i.e., boundary-adjacent “edge” cells versus those one or two celldiameters away (hereafter −1 and −2, respectively), and so on(see Fig. 3). Most WT axial mesoderm cells at midgastrulation(80% epiboly) were largely ML oriented and somewhat elongated,with boundary-adjacent “edge” cells being the least well oriented(median angle= 24.6°) (Fig. 3a–d). However, at the end of thegastrula period 90min later, edge cells became highly aligned andelongated (median angle= 13.6°) similar to internal cell rows(Fig. 3a–d). All MZudu−/− axial mesoderm cells exhibited sig-nificantly reduced ML alignment (Fig. 3g, Kolmogorov–Smirnovtest p < 0.0001), and elongation (Fig. 3h, Mann–Whitney testp < 0.0001) at 80% epiboly, but 90 min later only the edge cellsremained less aligned than WT (median angle= 17.0°) (Fig. 3g),although AR of the edge and −1 cells remained reduced (Fig. 3h).These results indicate significantly reduced ML orientation ofaxial mesoderm cells in MZudu−/− gastrulae, a defect thatpersisted only in boundary-adjacent cells at late gastrulation.Importantly, this reduction in ML cell polarity was accompaniedby significantly fewer cell intercalation events within the axial

mesoderm of MZudu mutants compared to WT (Fig. 3i–k, T-testp < 0.05). As ML intercalation is the key cellular behaviorrequired for vertebrate C&E3,5, we conclude that this is likelythe primary cause of axial mesoderm extension defects inMZudu mutants. We also observed significantly fewer mitoses inMZudu−/− gastrulae (Fig. 3l–n, T-tests), consistent with reportsof Zudu mutants28. Because decreased cell proliferation and theresulting reduced number of axial mesoderm cells were demon-strated to impair extension in zebrafish31, this could also be acontributing factor. Together, reduction of ML cell polarity, MLcell intercalations, and cell proliferation provide a suite ofmechanisms resulting in impaired axial extension in MZudu−/−gastrulae. Combined with irregular notochord boundariesobserved in MZudu mutants, we hypothesize that this boundaryprovides a ML orientation cue that contributes to ML polarizationof axial mesoderm cells, and that this cue is absent or reduced inMZudu−/− gastrulae.

Gon4l functions independently of PCP signaling. Molecularcontrol of ML cell polarity underlying C&E movements invertebrate embryos is chiefly attributed to PCP and Gα12/13signaling6–10. While zygotic loss of udu enhanced axial extensiondefects in knym818/m818 PCP mutants (Fig. 1), it was unclearwhether udu functions within or parallel to the PCP network.To address this, we generated compound MZudu;Zknyfr6/fr6

(a nonsense/null kny allele8) mutants utilizing germline replace-ment as described above. Strikingly, these compound mutant

MZudu N = 43

WT N = 23[****

Boundary straightnessMZuduWT

WT

MZ

udu

80% epiboly +30 min +60 min +90 min +170 min

N = 4 N = 6

WT MZudu

α Laminin

f g

a

b

c d 1.3

1.2

Tota

l / n

et b

ound

ary

leng

th

1.1

1.0

80%

+30

min

+60

min

+90

min

e

Fig. 2MZudumutant gastrulae exhibit irregular notochord boundaries. a–b Still images from live Nomarski time-lapse series of the dorsal mesoderm in WT(a) and MZudu−/− embryos (b) at the time points indicated. Images are representative of over 40 MZudu−/− gastrulae. Arrowheads indicate notochordboundaries. c–d Live confocal microscope images of representative WT (c, N= 23) and MZudu−/− (d, N= 43) embryos expressing membrane Cherry.Yellow lines mark notochord boundaries. e Quantification of notochord boundary straightness in live WT and MZudu−/− gastrulae throughoutgastrulation. Symbols are means with SEM (two-way ANOVA, ****p < 0.0001). f–g Confocal microscope images of immunofluorescent staining forpan-Laminin in WT (f) and MZudu−/− (g) embryos at two-somite stage. N indicates the number of embryos analyzed. Scale bars are 50 μm. Anterior isup in all images




embryos were substantially shorter than single MZudu orknyfr6/fr6 mutants (also strict maternal udu mutants, whichhave no obvious phenotype) (Fig. 4a–d). Likewise, interferencewith vangl2/tri function in MZudu mutants by injection ofMO1-vangl2 antisense morpholino oligonucleotide (MO)32 alsoexacerbated axis extension defects of MZudu mutants (Supple-mentary Fig. 4a-d). That reduced levels of PCP components kny

or tri enhanced short axis phenotypes resulting from completeudu deficiency provides evidence that Gon4l affects axial exten-sion via a parallel pathway. Furthermore, expression domains ofgenes encoding Wnt/PCP signaling components kny, tri, andwnt5 in MZudu−/− gastrulae were comparable to WT (Supple-mentary Fig. 4e-l). Finally, we found that the asymmetric intra-cellular localization of Prickle (Pk)-GFP, a core PCP component,

**** **** **** *** **** ******** **** ****

Edge

80% Epiboly 80% Epiboly+90 min +90 min

80% Epiboly +90 min

N = 32

Inte

rcal

atio

nsCell orientation Cell elongation80% Epiboly +90 min

N = 19

WT

MZ

udu

N = 4 ea

80% epiboly

α pH

3

WT MZudu

i j

l m

e f g h

c da b 100

80

60

Deg

rees

from

M-L

40

20

0

100

80

60

Deg

rees

from

M-L

40

20

WT MZudu WT MZudu WT MZudu

Gain Loss Total

80% TB 2SS 6SS

Axial mesoderm cell intercalationsk

n

600

* *

*

400

T1

exch

ange

s ov

er 9

0 m

in

200

0

400Cell proliferation

**** ****

***

*

pH3+

cel

ls/e

mbr

yo 300

200

100

WT

WT

MZ

udu

MZ

udu

WT

MZ

udu

WT

MZ

udu

0

0

7

6

5

Asp

ect r

atio

4

3

2

1

0

7

6

5

Asp

ect r

atio

4

3

2

1

0

–1 –2 Edge –1 –2–3 Edge –1 –2 Edge –1 –2–3

Edge –1 –2 Edge –1 –2–3 Edge –1 –2 Edge –1 –2–3

Fig. 3Mediolateral cell polarity and cell intercalations are reduced in the axial mesoderm of MZudu−/− gastrulae. a–b, e–f Still images from live time-lapseconfocal movies of the axial mesoderm in WT (a, b) and MZudu−/− (e, f) gastrulae at the time points indicated. Cell outlines are colored according to acell’s position with respect to the notochord boundary. c, g Quantification of axial mesoderm cell orientation at 80% epiboly (left side) and +90min (rightside) time points. Each dot represents the orientation of the major axis of a single cell with respect to the embryonic ML axis and is colored according tothat cell’s position with respect to the notochord boundary (as in images to the left). Bars indicate median values. Asterisks indicate significant differencesbetween WT and MZudu−/− (Kolmogorov–Smirnov test, ***p < 0.001, ****p < 0.0001). d, h Quantification of axial mesoderm cell elongation at 80%epiboly (left side) and +90min (right side) time points. Each dot represents the aspect ratio of a single cell and is color-coded as in c. Bars indicate meanvalues. Asterisks indicate significant differences between WT and MZudu−/− (Mann–Whitney test, ****p < 0.0001). i–j Cell intercalations detected in theaxial mesoderm of WT (i) and MZudu−/− gastrulae (j). Cells gaining contacts with neighbors are green, cells losing contacts are magenta, and cells thatboth gain and lose contacts are yellow. k Quantification of cell intercalation events (T1 exchanges) in WT (blue bars) and MZudu−/− gastrulae (greenbars) over 90min. N indicates the number of embryos analyzed. Bars are means with SEM (T-test, *p < 0.05). l–m 200 μm confocal Z projections ofimmunofluorescent staining for phosphorylated Histone H3 (pH3) in WT (l) and MZudu−/− gastrulae (m) at 80% epiboly. Images are representative ofeight independent trials. n Quantification of pH3+ cells/embryo at the stages indicated. Each dot represents a single embryo, dark lines are means withSEM (T-test, *p < 0.05, ***p < 0.001, ****p < 0.0001). Scale bars are 50 μm. Anterior/animal pole is up in all images





to anterior cell membranes in WT gastrulae during C&Emovements33,34 was not affected in MZudu−/− gastrulae(Fig. 4e, f). This is consistent with intact PCP signaling in MZudumutant gastrulae, and provides further evidence that Gon4lfunctions largely in parallel to PCP to regulate ML cell polarityand axis extension during gastrulation.

Gon4l-dependent boundary cue cooperates with PCP signaling.In addition to PCP signaling, notochord boundaries are requiredfor proper C&E of the axial mesoderm in ascidian embryos35 andare involved in the polarization of intercalating cells duringXenopus gastrulation1. Boundary defects observed in MZudu−/−gastrulae are not a common feature of mutants with reducedC&E, however, as notochord boundaries in knyfr6/fr6 gastrulaewere straighter than in WT (Fig. 4g). Consistent with cell polaritydefects previously reported in kny mutants8, all axial mesodermcells failed to align ML within kny−/− embryos at 80% epibolyregardless of their position relative to the notochord boundary(Fig. 4h–k). After 90 min, however, kny−/− cells in the edge (and

to a lesser extent, −1) position attained distinct ML orientation(median angle= 22.1°)(Fig. 4j). Indeed, the nearer a kny−/− cellwas to the notochord boundary, the more ML aligned it was likelyto be. This suggests that the notochord boundary provides a MLorientation cue that is independent of PCP signaling and func-tions across approximately two cell diameters. Furthermore, thisboundary-associated cue appears to operate later in gastrulation,whereas PCP-dependent cell polarization is evident by 80% epi-boly. Importantly, distinct cell polarity phenotypes observed inMZudu−/− and kny−/− gastrulae provide further evidence thatGon4l functions in parallel to PCP signaling.

To assess how PCP signaling interacts with the proposed Gon4l-dependent boundary cue, we examined the polarity of axialmesoderm cells in compound zygotic kny−/−;udu−/− mutantgastrulae (Fig. 4l–o). As observed in single kny−/− mutantgastrulae8 (Fig. 4h–k), both ML orientation and elongation of allaxial mesoderm cells were reduced in double mutants at 80%epiboly, regardless of position with respect to the notochordboundary (Fig. 4n, o). By late gastrulation, however, edge cells indouble mutant gastrulae failed to attain the ML alignment observed

kny–/–

Mudu sibling

WT N = 23 kny–/– N = 15

***

9/29 14/29

3/29 3/29

80% Epiboly 80% Epiboly+90 min +90 min

80% Epiboly +90 min

N = 6

Edge –1 –2 –3

Edge –1 –2 –3–4 –5

Edge –1 –2 –3

Edge –1 –2 –3–4 –5

Edge –1 –2 –3

Edge –1 –2 –3–4 –5

Edge –1 –2 –3

Edge –1 –2 –3–4 –5

Boundary straightness

Cell orientation Cell elongation80% Epiboly +90 min

kny–

/–kn

y–/–

; udu

–/–

N = 24

WT MZudua b

c d

e g

h i j k

l m n o

Pk-GFP

MZudu–/–;kny–/–Pk-GFP localization

N = 957 cells

N = 323 cells

WT MZudu

f100 Cytoplasmic

Other punctaAnterior puncta80

6040200

% c

ells

100

80

60

40

20

0

Deg

rees

from

M-L

100 76543210

Asp

ect r

atio

76543210

Asp

ect r

atio

80

60

40

20

0

Deg

rees

from

M-L

MZudu–/–1.3

****1.2

1.1

1.0

80%

+30

min

+60

min

+90

min

Tot

al /

net b

ound

ary

leng

th

Fig. 4 Gon4l regulates axis extension independent of PCP signaling. a–d Live embryos at 24 hpf resulting from a cross between a germline-replacedudu−/−;kny fr6/+ female and an udu+/-;knyfr6/+ male. Genotypes are indicated in the upper right corner, fractions indicate the number of embryos in theclutch with the pictured phenotype. eMosaically expressed Prickle (Pk)-GFP in WT and MZudu−/− gastrulae. Arrowheads indicate anteriorly localized Pk-GFP puncta. Membrane Cherry marks cell membranes, nuclear-RFP marks cells injected with pk-gfp RNA. Images are representative of three independentexperiments. f Quantification of Pk-GFP localization shown in e (chi-square, p= 0.07). g Quantification of notochord boundary straightness in WT andkny−/− gastrulae. Symbols are means with SEM (two-way ANOVA, ****p < 0.0001). h, i, l, m Still images from live time-lapse confocal movies of the axialmesoderm in kny−/− (h–i) and kny−/−;udu−/− (l–m) gastrulae at the time points indicated. Cell outlines are colored as in Fig. 3. j, n Quantification ofaxial mesoderm cell orientation as in Fig. 3, bars are median values (Kolmogorov–Smirnov test, ***p= 0.0003). k, o Quantification of axial mesodermcell elongation as in Fig. 3, bars are mean values. N indicates number of embryos analyzed (and number of cells in f). Scale bar is 500 μm in a–d, 10 μm in e,50 μm in h–m




in kny−/− mutants, but instead remained largely randomlyoriented (median angle= 38.0°) (Fig. 4n). This exacerbation of cellorientation defects correlated with stronger axis extension defects incompound kny−/−;udu−/− compared to single kny−/− mutants(Fig. 1e). This supports our hypothesis that a Gon4l-dependentboundary cue regulates ML alignment of axial mesoderm cellsindependent of PCP signaling, and that these two mechanismspartially overlap in time and space to cooperatively polarize all axialmesoderm cells and promote axial extension.

Loss of Gon4l results in large-scale gene expression changes. Asa nuclear-localized chromatin factor18,36 (Supplementary Fig. 3j),Gon4l is unlikely to influence morphogenesis directly. To identifygenes regulated by Gon4l with potential roles in morphogenesis,we performed RNA sequencing (RNA-seq) in MZudu−/− andWT tailbud-stage embryos. Analysis of relative transcript levelsrevealed that more than 11% of the genome was differentiallyexpressed (padj < 0.05, ≥twofold change) in MZudu mutants(Fig. 5a). Of these ~2950 differentially expressed genes, 1692exhibited increased expression in MZudu mutants compared to1259 with decreased expression (Supplementary data 1). Func-tional annotation analysis revealed that genes downregulated inMZudu−/− gastrulae were enriched for ontology terms related tochromatin structure, transcription, and translation (Fig. 5b, c),while upregulated genes were enriched for terms related to bio-synthesis, metabolism, and protein modifications (Fig. 5b, d).Notably, many of these misregulated genes are considered to have“housekeeping” functions, in that they are essential for cellsurvival. We further examined genes with plausible roles inmorphogenesis, including those encoding signaling and adhesionmolecules, and found the majority of genes in both classes wereexpressed at higher levels in MZudu−/− gastrulae (Fig. 5e, f).This putative increase in adhesion was of particular interest giventhe tissue boundary defects observed in MZudu mutants.

DamID-seq identifies putative direct targets of Gon4l. Todetermine genomic loci with which Gon4l protein associates, andthereby distinguish direct from indirect targets of Gon4l regula-tion, we employed DNA adenine methyltransferase (Dam)identification paired with next generation sequencing (DamID-seq)20,21. To this end, we generated Gon4l fused to Escherichiacoli Dam, which methylates adenine residues within genomicregions in its close proximity20. Small equimolar amounts of RNAencoding a Myc-tagged Gon4l-Dam fusion or a Myc-tagged GFP-Dam control were injected into one-celled embryos, and genomicDNA was collected at tailbud stage (see Methods section). Insupport of this Gon4l-Dam fusion being functional, it localized tothe nuclei of zebrafish embryos and partially rescued MZudu−/−embryonic phenotypes (Supplementary Fig. 5a-k). Methylated,and therefore Gon4l-proximal, genomic regions were thenselected using methylation specific restriction enzymes andadaptors, amplified to produce libraries, and sequenced. BecauseDam is highly active, even an untethered version methylates DNAwithin open chromatin regions20, and so libraries generated fromembryos expressing GFP-Dam served as controls.

Unbiased genome-wide analysis of DamID reads revealed asignificant enrichment of Gon4l-Dam over GFP-Dam inpromoter regions, and a significant underrepresentation of Gon4lwithin intergenic regions (Fig. 6a, T-test p < 0.05, Fig. 6b).Although no global difference was detected within gene bodies(Fig. 6a), examination of individual loci revealed ~4500 genes andover 2300 promoters in which at least one region was highlyGon4l-enriched (Padj ≤ 0.01, ≥fourfold enrichment over GFP)(Fig. 6c–g, Supplementary Data 2, 3). Of these, ~1000 genes wereco-enriched for Gon4l in both the promoter and gene body

(Fig. 6c, g). Levels of Gon4l association across a gene and itspromoter were significantly correlated (Supplementary Fig. 5l,Spearmann correlation p < 0.0001), indicating co-enrichment orco-depletion for Gon4l at both gene features of many loci. Withingene bodies, we found robust enrichment specifically within 5′untranslated regions (UTRs) (Fig. 6b), consistent with associationof Gon4l at or near transcription start sites (Fig. 6d). Of the~2950 genes differentially expressed in MZudu−/− gastrulae,approximately 28% (812) were also enriched for Gon4l at thegene body (492), promoter (170), or both (150) (Fig. 6g), and willhereafter be described as putative direct Gon4l targets. histh1, forexample, was among the most downregulated genes by RNA-seqin MZudu mutants (Supplementary Data 1) and was highlyenriched for Gon4l at both its promoter and gene body (Fig. 6c).By contrast, tbx6 expression was also reduced in MZudu mutants(Supplementary Data 1), but exhibited no enrichment of Gon4lover GFP controls (Fig. 6f). Approximately 35% and 53% of genesenriched for Gon4l in only the gene body or only the promoter,respectively, were positively regulated (i.e., downregulated inMZudu mutants), as were 50% of genes co-enriched at bothfeatures (Fig. 6h). Furthermore, among these positively regulatedgenes, differential expression levels (the degree to which WTexpression exceeded MZudu−/−) correlated positively andsignificantly with Gon4l-enrichment levels across both genebodies and promoters (Fig. 6h, Spearmann correlation p < 0.01).A similar correlation was not observed for negatively regulatedgenes, hence, the highest levels of Gon4l enrichment wereassociated with positive regulation of gene expression. Theseresults implicate Gon4l as both a positive and negative regulatorof gene expression during zebrafish gastrulation.

Gon4l limits itga3b to promote boundary straightness. We nextexamined our list of putative direct Gon4l target genes for thosewith potential roles in tissue boundary formation and/or cellpolarity. epcam, which encodes Epithelial cell adhesion molecule(EpCAM), stood out because it was not only enriched for Gon4lby DamID (Fig. 7a) and upregulated in MZudu−/− gastrulae(Fig. 5f, Supplementary Data 1), but was also identified ina Xenopus overexpression screen for molecules that disrupttissue boundaries37. Furthermore, EpCAM negatively regulatesnon-muscle myosin activity38, making it a compelling candidate.We also chose to examine itga3b, which encodes Integrinα3b,because as a component of a laminin receptor39, it is anobvious candidate for a molecule involved in formation of atissue boundary at which laminin is highly enriched (Fig. 2f).DamID revealed a region within the itga3b promoter at whichGon4l was highly enriched (Fig. 7b, c), and itga3b expressionwas increased in MZudu−/− gastrulae by RNA-seq (Supple-mentary Data 1), which was also validated (along with epcam)by qRT-PCR (Fig. 7d, e). We found that overexpression ofeither epcam or itga3b by RNA injection into WT embryosrecapitulated the irregular notochord boundaries (Fig. 7f, g)and reduced ML orientation and elongation of axial mesodermcells (Supplementary Figs. 6d-g, 7c-f) seen in MZudu−/−gastrulae. Conversely, while injection of an epcam MO (MO2-epcam40, Supplementary Fig. 7b) had no effect on MZudu−/−boundaries (Fig. 7h), injection of a translation-blocking itga3bMO (MO1-itga3b41, Supplementary Fig. 6b) significantlyimproved boundary straightness in MZudu mutants comparedwith control-injected siblings (Fig. 7i, two-way ANOVAp < 0.0001). This indicates that excess itga3b is largely responsiblefor reduced straightness of notochord boundaries in MZudu−/−gastrulae. Furthermore, disruption of WT notochord boundariesby excess itga3b was abrogated by co-injection with a MO againstlama5 (MO3-lama542, Supplementary Fig. 6c) (Fig. 7j), which





encodes its ligand Lamininα541, indicating that this effect isligand dependent. Both itga3b and lama5 are expressed within theaxial mesoderm of WT zebrafish gastrulae43,44. Moreover,although an Itgα3b-GFP fusion was not enriched at WT noto-chord boundaries, it became increasingly localized to the plasma

membrane of axial mesoderm cells (but not presomitic mesodermor neuroectodermal cells) as gastrulation proceeded (Fig. 7k–m).This implies that intracellular localization (and likely function) ofIntegrinα3b is regulated in a stage- and tissue-dependent manner,as was reported for other Integrins during Xenopus gastrulation45.

Calcium-binding protein

Cell adhesion molecule

Cell junction protein

Chaperone

Cytoskeletal protein

Defense/immunity protein

Enzyme modulator

Extracellular matrix protein

Hydrolase

Isomerase

Ligase

Lyase

Membrane traffic protein

Nucleic acid binding

Oxidoreductase

Receptor

Signaling molecule

Storage protein

Structural protein

Transcription factor

Transfer/carrier protein

Transferase

Transporter

Transmembrane receptor adaptor protein

Viral protein

Protein class:

2.4%

5.7%

6.5%

5.6%

5.5%

8.2%5%

2.4%

0.8%2.2%

12.6%

23.5%

3.4%

2.8%

Decreased expression in MZudu

Increased expression in MZudu

Fol

d ch

ange

WT

/MZ

udu

(log2

)

Mean expression

WT versus MZudu gene expression at tailbud stage

Decreased expression in MZudu

Increased expression in MZudu

Color key

WT MZudu

3

2

1

0

–1

–2

–3

WT MZudu

Signaling molecules Adhesion moleculesc

d

e

a b

1e+01 1e+02 1e+03 1e+04 1e+05

DNA bindingNucleic acid binding

Nucleosome assembly

Oligosaccharide biosynthesisNeuropilin binding

RNA bindingHistone binding

Ribosome biogenesisrRNA processing

Regulation of transcriptionChromatin binding

Helicase activity

100 10–5 10–10 10–15 10–20

P -value

P -value

jam2aepcamtspan36tgfbicd63cd81akitlgatspan37he1btspan2bIgals1l1cxcr4bIgals2azgc:110329tspan15pcdh10aigdcc4sdc4cd9almlntmem8aigsf9bbpcdh8cdcp2si:ch211–74m13.3chl1apostnbmyom3cdh5cntnap2acrim1chl1bdscaml1ccdc141ltbp1cubnBX511034.2mxra5anexncdh2tspan4aprtgaccr6asi:dkey–11f4.20

–2 –1 0 1 2Row Z -score

icn2rcan1aicnsst6s100v2tspan36npyddttgfbiangptl2bcd63cops8sema3blcd81azgc:92313fxd2fgbkitlgatspan37he1befcab11namptatspan2bzgc:66350adcyap1algals1l1cbln12f10lgals2atraf6wnt11rpnocbinhbbcalcaldlrap1azgc:110329plxna4il6sttspan15SEMA4Fsdc4gstt1asema3ecd9anos1apatraf4bdab2ppp1r1belmod3il1bdvl2platarhgef25bil6ril12rb2fli1asema4bacdcp2si:ch211–74m13.3ppp1r1cpostnbhbegfbrcan3fndc7rs4plgsema4gacishbsema3abwnt4bsema6bbDAB2cntnap2atncItbp1tmprss9cubnnrg2bloxhd1ahmgxb4achgaubtfkalrnbg3bp1ikplekhg4ubtflmarcksl1bwnt5bergplxna1btnfsf10l3etv5ba2mlloxhd1btspan4aeef1gcyr61l1lsg1hmgb1ahmgb3asema6dlft2ssrp1andr2hmgb2ahdgfrp2s100thmgb2bil23rsi:dkey–11f4.20

Metabolic process

Semaphorin receptor bindingTransferase activity

Hydrolase activityOxidation-reduction process

Negative chemotaxisCatalytic activity

Oxidoreductase activity

0.00

001

0.00

010.

001

0.010.

11

f

Fig. 5 Loss of Gon4l results in large-scale gene expression changes. a Plot of gene expression changes in MZudumutants compared to WT at tailbud stageas assessed by RNA sequencing. Red dots indicate genes expressed at significantly different levels than WT (p≤ 0.05, at least twofold change in transcriptlevel). b Protein classes encoded by differentially expressed genes with decreased (top graph) or increased (bottom graph) expression in MZudu mutants.Percentages indicate the number of genes within a given class/total number of genes with increased or decreased expression, respectively. c–d P-values ofthe top ten most enriched gene ontology (GO) terms among genes with decreased (c) or increased (d) expression in MZudu−/− gastrulae compared toWT. e–f Heat maps of differentially expressed genes annotated as encoding signaling (e) or adhesion molecules (f). The four columns represent twobiological and two technical replicates for each WT and MZudu−/−




Finally, to determine whether excess itga3b was sufficient tophenocopy the kny−/−;udu−/− double mutant phenotypeidentified in our synthetic screen (Fig. 1), we overexpressed itga3bin kny−/− embryos and found that it exacerbated the short axisphenotype of these PCP mutants (Fig. 7n–p), an effect not pro-duced by injection of control GFP RNA. Together, these resultsindicate that negative regulation of itga3b expression by Gon4l isessential for proper notochord boundary formation and axisextension in zebrafish gastrulae. Interestingly, ML cell polaritydefects were not suppressed in MZudu−/− embryos injected withitga3b (or epcam) MO compared to control-injected mutantsiblings (Supplementary Figs. 6h-o, 7g-n), despite improved

boundary straightness. This implies that otherGon4l-dependent boundary properties are involved, and/or thatloss of udu affects ML polarity cell-autonomously, for example,by making cells unable to respond to the boundary-associatedpolarity cue.

Loss of Gon4l and excess epcam reduce boundary tension. InWT zebrafish and Xenopus embryos, notochord boundariesstraighten over time (Fig. 2e) and accumulate myosin46, implyingthat they are under tension. Because boundaries of MZudumutants are less straight than those of WT gastrulae, we hypo-thesized that tension at these boundaries is reduced. To test this,

3009 1201

2138

854

170492

150

p = 0.013

p = 0.026

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

GFP-Dam

Gon4l-Dam

GFP-Dam

Gon4l-Dam

GFP-Dam

Gon4l-Dam

Gon

4l-D

am

Gon4l/GFP uniquely aligned reads

GF

P-D

am

Scale = 0–64

14 kb

tbx6

2491 bp

Scale = 0–1012

cldn5

22 kb

Scale = 0–416

wnt11r

Gon

4/G

FP

uniq

uely

alig

ned

read

s(L

og2

fold

cha

nge)

3

2

1

0

–1

Inte

rgen

ic

Prom

oter

5′ UTR

3′ UTR

ORF

5

4

3

2

1

0

–1

–2

–3

–4

–52 4 6 8 10

Gon4l enrichment (log2 fold change over GFP)

Gon4l enrichment at genes and promotersversus differential expression

Diff

eren

tial g

ene

expr

essi

on(lo

g2 fo

ld c

hang

e W

T/M

Zud

u)

2992 bp

Scale = 0–4602

histh1

GF

P-D

amG

on4l

-Dam

Gene bodies

Promoters

Intergenicregions

RPKM of uniquely aligned reads

Gon4l-enrichedpromoters

Gon4l-enrichedgene bodies

Differentially expressedin MZudu–/–

No Gon4l enrichment

Gon4l-enriched Gon4l-enriched

Gon4l-enriched

Gon4l-enriched promotersGon4l-enriched genes

R = 0.20**R = 0.16**

a

b

g h

c

e

d

f

Fig. 6 DamID-seq identifies putative direct targets of Gon4l during gastrulation. a Box plot of normalized uniquely aligned DamID reads at each of threecategories of genomic regions. Center lines are medians, box limits are upper and lower quartiles, whiskers are highest and lowest values. P-values indicatesignificant differences between Gon4l-Dam samples and GFP-Dam controls (T-test). b Fold change (log2) Gon4l-Dam over GFP-Dam reads (RPKM) ateach of five gene features. c–f Genome browser tracks of Gon4l-Dam and GFP-Dam association at the histh1 (c), wnt11r (d), cldn5 (e), and tbx6 (f) loci.Scale is number of reads. Each track represents one biological replicate at tailbud stage. g Venn diagram of genes with regions of significant Gon4lenrichment within gene bodies (blue) or promoters (pink), and genes differentially expressed in MZudu−/− gastrulae (gray). h Correlation of Gon4lenrichment levels across gene bodies (blue dots) and promoters (red dots) with relative transcript levels of genes differentially expressed in MZudumutants. A positive correlation was detected between increased expression in WT relative to MZudu mutants and Gon4l enrichment in both gene bodies(Spearman correlation p= 0.0095) and promoters (p= 0.0098). Dotted lines are linear regressions of these correlations





we laser-ablated interfaces between axial mesoderm cells in liveWT and MZudu−/− gastrulae and recorded recoil of adjacentcell vertices as a measure of tissue tension47 (Fig. 8a, b). Cellinterfaces were classified according to an established convention

as V junctions (actively shrinking to promote cell intercalation), Tjunctions (not shrinking)47,48 or Edge junctions (falling on andthus comprising the notochord boundary). We found that recoildistances at WT Edge junctions were significantly greater than

[****

[****

[****

[****

Scale = 0–146

4500 bp

epcam

700 bp

Scale = 0–150

N = 2315

N = 23

16N = 9

119

N = 15

11

N = 8

10

itga3b

85 kb




Boundary straightness Boundary straightness

kny–/– +GFP

24 hpf

kny–/– +itga3b

Axi

s st

reng

th a

t 24

h(m

m)

kny–

/–

kny–

/–

+200

itga3

b

kny–

/–

+200

GFP

Gon

4l-D

amG

FP

-Dam

epcam expression at TB itga3b expression at TB

Axial meso 80% Axial meso +90 min Neural +90 min

Itgα3

b-G

FP

N = 6N = 2 N = 3 N = 7

N = 14

n o p

mlk

jig

hf

d ecba

5

4

1.3MZudu

WTWT +200 epcam

MZudu +epcam MO

MZudu sibs

WT sibs

WT1.2

Tot

al/n

et b

ound

ary

leng

th

1.1

1.0

1.3

1.2

Tot

al/n

et b

ound

ary

leng

th

1.1

1.0

80%

+30

min

+60

min

+90

min

1.3

MZudu

WT

WT +200 itga3b

WT +itga3b +lama5 MOWT +200 itga3b

1.2

Tot

al/n

et b

ound

ary

leng

th

1.1

1.0

80%

+30

min

+60

min

+90

min

1.3

1.2

Tot

al/n

et b

ound

ary

leng

th

1.1

1.0

80%

+30

min

+60

min

+90

min

1.3

1.2

Tot

al/n

et b

ound

ary

leng

th

1.1

1.0

3 *******NS

N = 20 N = 27 N = 68

2

1

0

80%

+30

min

+60

min

+90

min

80%

+30

min

+60

min

+90

min

3

Fol

d ch

ange

2

1

0

5

4

3

Fol

d ch

ange

2

1

0WT MZudu WT MZudu

MZudu +itga3b MO

MZudu sibs

WT




that of WT V (Kruskal–Wallis test p= 0.0001) or T junctions(Kruskal–Wallis test p= 0.0026), demonstrating that the noto-chord boundary is under greater tension than the rest of the tissue(Fig. 8c–e). Consistent with our hypothesis, recoil distance wassignificantly smaller in MZudu−/− than in WT gastrulae for allclasses of junctions (Fig. 8c–e, two-way ANOVA p < 0.01), andthis decrease was largest and most significant in Edge junctions(Fig. 8c, two-way ANOVA p < 0.0001).

Because excess epcam and itga3b were sufficient to reduceboundary straightness (Fig. 7), we next tested whether epcam anditga3b overexpression could likewise affect boundary tension.Injection of WT embryos with epcam RNA was sufficient to lowertissue tension at the notochord boundary and throughout theaxial mesoderm (Fig. 8c–e). Curiously, although we found itga3bto be largely causative of reduced boundary straightness inMZudu mutants (Fig. 7), its overexpression did not produce asimilar reduction in tension (Fig. 8c–e), implying a tension-independent role for Itgα3b in boundary straightness. Accord-ingly, treatment of itga3b-overexpressing WT embryos withCalyculin A, a myosin phosphatase inhibitor that increasesmyosin contractility49 and notochord boundary tension (Supple-mentary Fig. 8a), did not restore their boundary straightnessdespite increased tension (Fig. 8f, Supplementary Fig. 8c). Bycontrast, Calyculin A treatment significantly increased boundarytension and straightness in epcam-overexpressing WT embryos(Fig. 8g, Supplementary Fig. 8d, two-way ANOVA), indicating atension-dependent role for EpCAM in boundary formation.Consistent with its previously described inhibition of myosinactivity38, these data support a role for EpCAM as a negativeregulator of myosin-dependent tissue tension at the notochordboundary. Notably, treatment of MZudu mutants with CalyculinA did not restore boundary straightness or axial mesoderm cellpolarity (Supplementary Fig. 8b, e-h), demonstrating thatincreasing tension alone was not sufficient to improve thesephenotypes. Given that the effect of EpCAM on the boundary islargely tension dependent, this likely explains why the epcam MOfailed to improve boundary straightness in MZudu mutants(Fig. 7h). Together these results implicate excess EpCAM andIntegrinα3b as key molecular defects underlying reducedboundary tension and reduced boundary straightness, respec-tively, observed in MZudu−/− gastrulae. We propose a modelwhereby Gon4l limits expression of these adhesion molecules toensure proper formation of the notochord boundary, whichtogether with additional boundary-independent roles of Gon4lcooperates with PCP signaling to promote ML cell polarityunderlying C&E gastrulation movements (Fig. 8h).

DiscussionSubstantial advances have been made in defining signalingpathways that regulate gastrulation cell behaviors and shape thevertebrate body plan6,7,9,10, but epigenetic control of these mor-phogenetic processes remains largely unexplored. Here we have

described the conserved chromatin factor Gon4l as a regulator ofpolarized cell behaviors underlying axis extension during zebra-fish gastrulation. We identified a large number of Gon4l targetgenes, including many with known or predicted roles in mor-phogenesis, and linked misregulation of a subset of these genes tospecific morphogenetic defects. Because Gon4l does not bindDNA directly, we predict that Gon4l-enriched genomic loci aredirect targets of chromatin modifying protein complexes withwhich Gon4l associates36. Only a fraction of the thousands ofGon4l-enriched genes and promoters exhibited correspondingchanges in gene expression during gastrulation, which may reflectone or more possible scenarios. Either Gon4l does not alterexpression of all loci with which it associates, loci recentlyoccupied by Gon4l do not yet reflect changes in transcript levels,or because DamID provides a “history” of Gon4l association, bothformerly and currently occupied loci are represented in our data.We also identified loci at which Gon4l was depleted compared toGFP-Dam controls (Supplementary Fig. 5), and speculate theyrepresent open chromatin regions with which Gon4l does notassociate. Surprisingly, although a larger number of putativeGon4l direct target genes were negatively regulated (i.e., upre-gulated in MZudu mutants), the highest levels of Gon4l enrich-ment were correlated with positive regulation of gene expression(Fig. 6). This argues against Gon4l acting strictly as a negativeregulator of gene expression, a role assigned to it based on in vitroevidence and thought to be mediated by its interactions withHistone deacetylases36. Our data instead indicate that Gon4l actsas both a positive and negative regulator of gene expressionduring zebrafish gastrulation, implying context-specific interac-tions with multiple epigenetic regulatory complexes.

Phenotypes caused by complete udu deficiency are con-spicuously pleiotropic, but our studies point to remarkably spe-cific roles for udu in gastrulation morphogenesis. Loss of udufunction reduced tissue extension without affecting mesendoderminternalization, epiboly, prechordal plate migration, or con-vergence (see below). Moreover, the dorsal gastrula organizer andall three germ layers were specified (Fig. 1, Supplementary Fig. 2),indicating that MZudumutants do not suffer from a general delayor arrest of development; rather Gon4l regulates a specific subsetof gastrulation cell behaviors, including ML cell polarity andintercalation in the axial mesoderm. Importantly, the role ofGon4l in these processes is independent of PCP signalingas supported by several lines of evidence (Fig. 4, SupplementaryFig. 4). Additionally, whereas PCP mutants exhibited reducedconvergence7,8, evidenced by a larger number of cell rows inkny−/− axial mesoderm (Fig. 4), MZudu mutants contained anormal number of axial mesoderm cell rows (Fig. 3), indicatingno obvious convergence defect. Despite these apparently parallelfunctions, RNA-seq and DamID-seq experiments revealed thatGon4l regulates expression of some PCP genes, including wnt11,wnt11r, prickle1a, prickle1b, celsr1b, celsr2, and fzd2 (Supple-mentary Data 1-3). However, functional redundancies within thePCP network50,51 likely allow for intact PCP signaling in MZudu

Fig. 7 Gon4l regulates notochord boundary straightness by limiting itga3b expression. a–c Genome browser tracks of Gon4l-Dam and GFP-Damassociation at the epcam (a) and itga3b (b) locus. An expanded view of the itga3b promoter is shown in c. d–e Quantitative RT-PCR for epcam (d) and itga3b(e) in WT and MZudu−/− embryos at tailbud stage. N indicates the number of independent clutches tested with technical triplicates of each, barsare means with SEM. f–g Quantification of notochord boundary straightness in WT epcam (f) or itga3b (g) overexpressing embryos (two-way ANOVA,****p < 0.0001). h–i Notochord boundary straightness in MZudu−/− epcam (h) or itga3b (i) morphants and sibling controls (two-way ANOVA,****p < 0.0001). j Quantification of notochord boundary straightness in WT itga3b-overexpressing embryos with or without lama5 MO (two-way ANOVA,****p < 0.0001). N indicates the number of embryos analyzed, symbols are means with SEM. k–m Itgα3b-GFP localization in WT embryos in the tissuesand at the time points indicated. Insets are enlarged from regions in white squares. Yellow lines and arrowheads mark notochord boundaries. Images arerepresentative of three independent trials. Scale bar is 50 μm. n–o Live images of kny−/− embryos at 24 hpf injected with 200 pg GFP (n) or 200 pg itga3b(o) RNA. Images are representative of four independent experiments. p Quantification of axis length of injected kny−/− embryos at 24 hpf. Each dotrepresents one embryo, bars are means with SEM (T-tests, ***p < 0.001, ****p < 0.0001). N indicates number of embryos analyzed





mutants despite misregulation of some PCP components. Nota-bly, gastrula morphology and cell polarity defects in MZudumutants were also distinct from ventralized and dorsalized pat-terning mutants with impaired C&E11,12.

Modulation of adhesion at tissue boundaries has been impli-cated as a driving force of cell intercalation46,52, and indeed genesannotated as encoding adhesion molecules tended to be expressed

at higher levels in MZudu−/− gastrulae (Fig. 5). Two of these,itga3b and epcam, exhibited increased expression in MZudu−/−gastrulae by RNA-seq and qRT-PCR, were identified as putativedirect Gon4l targets by DamID (Fig. 7), and were each sufficientto reduce notochord boundary straightness and ML cell polaritywhen overexpressed in WT embryos (Fig. 7, SupplementaryFigs. 6, 7). Furthermore, decreasing levels of Integrinα3b (but not

******

********

N = 36

N = 9

6

15 16

9

N = 9

34

32

40

Pos

t-ab

latio

nP

re-a

blat

ion

Edge V junx T junx

Edge recoil by time V junx recoil by time T junx recoil by time

Rec

oil d

ista

nce

(μm

)

2.5

V junx

T junx

Edge

Pre

Post

2.0

1.5

1.0

0.5

0.0

1.3

1.2

1.1

1.0

1.3

1.2

1.1

1.0

–0.5

80%

+30

min

+60

min

+90

min

80%

+30

min

+60

min

+90

min

80%

+30

min

+60

min

+90

min

80%

+30

min

+60

min

+90

min

80%

+30

min

+60

min

+90

min

2.5WT

WT cntrl WT cntrl

MZudu

WT+epcam

WT+itga3b

WT+itga3b

WT+itga3b +calA

WT+epcam

WT+epcam +calA

******

****

2.0

1.5

1.0

0.5

0.0

–0.5

2.5

2.0

1.5

1.0

0.5

0.0

–0.5

Rec

oil d

ista

nce

(μm

)

Rec

oil d

ista

nce

(μm

)

Tot

al/n

et b

ound

ary

leng

th

Tot

al/n

et b

ound

ary

leng

th

Boundary straightness Boundary straightness

a

g

b

c d

f

e

hGon4l

EpCAM

Ten

sion

Ten

sion

EpCAMItgα3b

Itgα3b

Laminin

Myosin

PCP signaling

Boundary cue

epcam/itga3b epcam/itga3b

WT WT

+90 min

MZudu–/– kny–/– kny–/–;udu–/–

80% epiboly




EpCAM) improved boundary straightness in MZudu mutants(Fig. 7), while overexpression of epcam (but not itga3b) wassufficient to reduce tissue tension at the notochord boundary ofWT embryos (Fig. 8). In addition, increasing tension with Caly-culin A in epcam-overexpressing (but not itga3b overexpressing)WT embryos normalized their notochord boundary straightness(Fig. 8). These observations support a model in which mis-regulation of each of these molecules contributes to distinct cel-lular defects in MZudu mutants: excess itga3b disrupts boundarystraightness in a tension-independent manner via interactionswith its laminin ligand, while EpCAM negatively regulatesmyosin-dependent boundary tension. EpCAM was similarlyshown to negatively regulate non-muscle myosin contractility inXenopus gastrulae38, where experimental perturbation of myosinactivity disrupts notochord boundary formation46.

Links between myosin-dependent tissue tension, boundaryformation, and cell intercalation are well established. In additionto Xenopus notochord boundaries46, myosin accumulationincreases tension at compartment boundaries in Drosophila53,which was shown to bias cell intercalations54. We found thatdecreased boundary straightness in epcam-overexpressing WTgastrulae could be rescued by pharmacologically increasing tissuetension, demonstrating a causal relationship between tension andboundary straightness (Fig. 8). However, increasing tissue tensionwas not sufficient to improve boundary straightness inMZudu−/− gastrulae (Supplementary Fig. 8), implying addi-tional molecular bases for this phenotype, namely excess itga3b.Furthermore, restoration of boundary straightness (with itga3bMO) in MZudu−/− gastrulae was not sufficient to normalize MLcell orientation, raising questions as to the functional relationshipbetween these elements in this context. We speculate thatincreased boundary straightness and tension together may berequired to promote ML polarity of boundary-adjacent cells.Alternatively, improved boundary straightness in MZudu−/−gastrulae may have restored the polarity cue, but additional geneexpression changes rendered mutant cells unable to respond to it.Although it is not yet clear whether and how boundarystraightness and tension affect the polarity of adjacent cells, evi-dence from this study and others strongly implicate the noto-chord boundary in ML polarization of cells undergoing C&E1,35.Namely, within kny−/− PCP mutants, only boundary-adjacentedge cells became ML aligned during late gastrulation. Togetherwith the additive cell polarity defects observed in compoundkny−/−;udu−/− mutants, this is strong genetic evidence that theboundary provides a Gon4l-dependent, PCP-independent cellpolarity cue (Fig. 4).

In this study of zebrafish Gon4l, we have begun to dissect thelogic of epigenetic regulation of gastrulation cell behaviors byrevealing a key role for this chromatin factor in limiting expres-sion of specific genes with specific morphogenetic consequences.This has important developmental implications, as C&E gas-trulation movements are sensitive to both gain and loss of genefunction7,10. We propose that by negatively regulating Integri-nα3b and EpCAM levels, Gon4l promotes development of theanteroposteriorly aligned notochord boundary and influences ML

polarity and intercalation of axial mesoderm cells. In cooperationwith PCP signaling, this cue coordinates ML cell polarity withembryonic patterning to drive AP embryonic axis extension(Fig. 8h).

MethodsZebrafish strains and embryo staging. Adult zebrafish were raised and main-tained according to established methods55 in compliance with standards estab-lished by the Washington University Animal Care and Use Committee. Embryoswere obtained from natural mating and staged according to morphology asdescribed56. All studies on WT were carried out in AB* or AB*/Tübingen back-grounds. Additional lines used include knym818, knyfr68, udusq118, and uduvu66 (thiswork). Embryos of these strains generated from heterozygous intercrosses weregenotyped by PCR after completion of each experiment. Germline-replaced fishwere generated by the method described in ref. 27. Briefly, donor embryos fromuduvu66/+ intercrosses or females with uduvu66/vu66 germline were injected withsynthetic RNA encoding GFP-nos1-3′UTR57, and WT host embryos were injectedwith a MO against dead end1 (MO1-dnd1)27 to eliminate host germ cells. Cellswere transplanted from the embryonic margin of donor blastulae to the embryonicmargin of hosts at sphere stage, and both hosts and donors were cultured inagarose-coated plates. Host embryos were screened for GFP+ germ cells at 36–48hpf, and the genotype of corresponding donors was determined by phenotype. Allputative uduvu66/vu66 germline hosts were raised to adulthood and confirmed bycrossing to uduvu66/+ animals prior to use in experiments. Fish were chosen fromtheir home tank to be crossed at random. The resulting embryos were also chosenfrom the dish at random for injection and inclusion in experiments. No fewer thansix (but often many more) embryos of each condition or genotype were analyzed inat least three independent trials for each experiment in which measurements weremade, except where noted. These rare cases reflect difficulty in obtaining sufficientMZudu−/− embryos due to limited availability and productivity of germline-replaced females.

Synthetic mutant screening. WT male fish were mutagenized by the chemicalmutagen ENU as described in ref. 22. Briefly, adult male fish were incubated four tosix times, for 1 h each, in a solution of 3.5 mM ENU at 21 °C. After treatment, fishwere allowed to recover for 2–3 h in 10 mg/L 3-aminobenzoic acid ethyl ester(MESAB) at 17–19 °C and an additional day without MESAB before returning tothe circulating water system. Approximately 3 weeks after mutagenesis, these maleswere outcrossed to WT females to produce F1 families. F2 families were obtainedby crossing F1 fish with fish homozygous for the hypomorphic knypek alleleknym818 (rescued by injection with synthetic kny WT RNA). F3 embryos obtainedfrom F2 cross were screened by morphology at 12 and 24 hpf to identify recessiveenhancers of the knym818/m818 short axis mutant phenotype8.

Positional cloning. We employed the positional cloning approach using a panel ofCA simple sequence length polymorphism markers representing 25 linkagegroups25 to map the vu66 mutation to chromosome 16 between the markers ofz17403 and z15431. Given phenotypic similarities between vu66/vu66 and theudusq1/sq1 mutant phenotype18, we sequenced udu cDNA from 24 hpf vu66/vu66mutant embryos revealing a T2261A transversion that is predicted to createY753STOP nonsense mutation. We designed a dCAPS58 marker for the vu66mutation and confirmed that no recombination occurred in 810 vu66/vu66homozygous embryos.

Microinjection. One-celled embryos were aligned within agarose troughs gener-ated using custom-made plastic molds and injected with 1–3 pL volumes usingpulled glass needles. Synthetic mRNAs for injection were made by in vitro tran-scription from linearized plasmid DNA templates using Invitrogen mMessagemMachine kits. Doses of RNA per embryo were as follows: 100 pg membraneCherry, 50 pg membrane GFP, 25 pg udu-gfp, 200 pg epcam, 200 pg itga3b, 50 pgitga3b-GFP, 1 pg gfp-dam-myc, 3 pg udu-dam-myc for DamID experiments, 20 pggfp-dam-myc for MZudu−/− rescue, and 150 pg GFP-nos1-3′UTR for germlinetransplantation. To assess Pk-GFP localization, embryos were injected at one-cellstage with membrane Cherry RNA, then injected with 15 pg Drosophila prickle-GFPand 20 pg H2B-RFP RNAs into a single blastomere at 16-cell stage as

Fig. 8 Loss of Gon4l and excess epcam reduce notochord boundary tension. a Diagram of laser ablation experiments to measure tension at axial mesodermcell interfaces. b Still images from confocal time-lapse movies of each of the three types of cell interfaces (Edge, V junctions, and T junctions) beforeand after laser ablation. Arrowheads indicate cell vertices adjacent to the ablated interface. Images are representative of 36 independent experiments.c–e Quantification of cell vertex recoil distance immediately after laser ablation of Edge (c), V junction (d), and T junction (e) interfaces at the time pointsindicated in WT, MZudu−/−, WT epcam overexpressing, and WT itga3b-overexpressing gastrulae. Symbols are means with SEM. Asterisks are coloredaccording to key and indicate significant differences compared to WT controls (two-way ANOVA, **** p < 0.0001, **p < 0.01). f–g Quantification ofnotochord boundary straightness in WT itga3b (f) or epcam (g) overexpressing embryos with or without Calyculin A (two-way ANOVA, ****p < 0.0001,***p < 0.001). N indicates the number of embryos analyzed, symbols are means with SEM. h Graphical model of the roles of Gon4l and PCP signaling inregulating ML cell polarity of axial mesoderm cells and notochord boundary formation





described33,34. Injections of MOs were carried out as for synthetic RNA. Doses ofMOs per embryo were as follows: 3 ng MO1-dnd127, 4 ng MO1-tri/vangl232, 1 ngMO2-epcam40, 2 ng MO1-itga3b41, and 1 ng MO3-lama542.

WISH. Antisense riboprobes were transcribed using NEB T7 or T3 RNA poly-merase and labeled with digoxygenin (DIG) (Roche). WISH was performedaccording to ref. 59. Briefly, embryos were fixed overnight in 4% paraformaldehyde(PFA) in phosphate buffered saline (PBS), rinsed in PBS+0.1% tween 20 (PBT),and dehydrated into methanol. Embryos were then rehydrated into PBT, incubatedfor at least 2 h in hybridization solution with 50% formamide (in 0.75M sodiumchloride, 75 mM sodium citrate, 0.1% tween 20, 50 μg/mL heparin (Sigma), and200 μg/mL tRNA) at 70 °C, then hybridized overnight at 70 °C with antisenseprobes diluted approximately 1 ng/μl in hybridization solution. Embryos werewashed gradually into 2X SSC buffer (0.3 M sodium chloride, 30 mM sodiumcitrate), and then gradually from SSC to PBT. Embryos were blocked at roomtemperature for several hours in PBT with 2% goat serum and 2mg/mL bovineserum albumin (BSA), then incubated overnight at 4 °C with anti-DIG antibody(Roche #11093274910) at 1:5000 in block. Embryos were rinsed extensively in PBT,and then in staining buffer (PBT+100 mM Tris pH 9.5, 50 mM MgCl2, and 100mM NaCl) prior to staining with BM Purple solution (Roche).

Immunofluorescent staining. Embryos were fixed in 4% PFA, rinsed in PBT,digested briefly in 10 μg/mL proteinase K, refixed in 4% PFA, rinsed in PBT, andblocked in 2 mg/mL BSA+ 2% goat serum in PBT. Embryos were then incubatedovernight in rabbit anti-Laminin (Sigma L9393) at 1:200, mouse anti-Myc (CellSignaling 2276) at 1:1000, or rabbit anti-phospho histone H3 (Upstate 06-570) at1:500 in blocking solution, rinsed in PBT, and incubated overnight in Alexa Fluor488 anti-Rabbit IgG, 568 anti-Rabbit, or 568 anti-Mouse (Invitrogen) at 1:1000 inPBT. Embryos were co-stained with 4′,6-Diamidino-2-Phenylindole, Dihy-drochloride and rinsed in PBT prior to mounting in agarose for confocal imaging.

TUNEL staining. Terminal deoxynucleotidyl transferase (TdT) dUTP Nick-EndLabeling (TUNEL) staining to detect apoptosis was carried out according to theinstructions for the ApopTag Peroxidase in situ apoptosis detection kit (Millipore)with modifications. Briefly, embryos were fixed in 4% PFA, digested with 10 μg/mLproteinase K, refixed with 4% PFA, and post-fixed in chilled ethanol:acetic acid 2:1,rinsing in PBT after each step. Embryos were incubated overnight with TdT andrinsed in stop/wash buffer, then blocked, incubated with anti-DIG antibody, andstained in Roche BM Purple staining solution.

Microscopy. Live embryos expressing fluorescent proteins or fixed embryos sub-jected to immunofluorescent staining were mounted in 0.75% low-melt agarose inglass bottomed 35-mm petri dishes for imaging using a modified Olympus IX81inverted spinning disc confocal microscope equipped with Voltran and Coboltsteady-state lasers and a Hamamatsu ImagEM EM CCD digital camera. For livetime-lapse series, 60 μm z-stacks with a 2 μm step were collected every three to tenminutes (depending on the experiment) for 3 h using a ×40 dry objective lens.Embryo temperature was maintained at 28.5 °C during imaging using a Live CellInstrument stage heater. When necessary, embryos were extracted from agaroseafter imaging for genotyping. For immunostained embryos, 200 μm z-stacks with a1 or 2 μm step were collected using a ×10 or ×20 dry objective lens, depending onthe experiment. Bright field and transmitted light images of live embryos andin situ hybridizations were collected using a Nikon AZ100 macroscope.

Laser ablation tension measurements. Embryos were injected at one-cell stagewith mCherry mRNA and mounted at 80% epiboly for imaging (as describedabove) on a Zeiss 880 Airyscan 2-photon inverted confocal microscope. Aninfrared laser tuned to 710–730 nm was used to ablate fluorescently labeled cellinterfaces, immediately followed by a quick time-lapse series of ten images with nointerval to record recoil after each ablation event. Images were collected using a ×40water immersion objective lens, and a Zeiss stage heater was used to maintainembryo temperature at 28.5 °C. Approximately 8–12 cell interfaces were ablatedper embryo. Image series were analyzed using ImageJ to determine the inter-vertexdistance of each cell interface prior to and immediately after ablation and used tocalculate recoil distance.

Calyculin A treatment. Embryos were mounted in agarose as described above,then embryo medium containing 50 nM Calyculin A (Sigma-Aldrich) was addedapproximately 30 min prior to the start of imaging and remained throughout theimaging period.

Image analysis. ImageJ was used to visualize and manipulate all microscopydatasets. For immunostained embryos, multiple z-planes were projected together tovisualize the entire region of interest. For live embryo analysis, a single z-planethrough the length of the axial mesoderm was chosen for each time point. Whenpossible, embryo images were analyzed prior to genotyping, and all images werecoded during analysis. To measure cell orientation and elongation, the AP axis inall embryo images was aligned prior to manual outlining of cells. A fit ellipse was

used to measure orientation of each cell’s major axis and its AR. The TissueA-nalyzer ImageJ package60 was used to automatically segment time-lapse series ofaxial mesoderm and detect T1 transitions. Boundary straightness was measured bymanually tracing the notochord boundary to determine total length, then dividingit by the length of a straight line connecting the ends of the boundary (net length).To assess Pk-GFP localization, isolated cells expressing Pk-GFP were scoredaccording to subcellular localization of GFP signal.

Quantitative RT-PCR. Total RNA was isolated from tailbud stage WT and MZudu−/− embryos homogenized in Trizol (Life Technologies), 1 μg of which was usedto synthesize cDNA using the iScript kit (BioRad) following manufacturer’s pro-tocol. SYBR green (BioRad) qRT-PCR reactions were run in a CFX Connect Real-Time PCR detection system (BioRad) in technical triplicate. Primers used are asfollows:

epcam: F-TGAGGACGGGGATTGAGAACR-GAGCCTGCCATCCTTGTCATitga3b: F-CCGGTGTTGGGAGAAGAGACR-CTTGAAGAAACCACACGAAGGGEF1a: F-CTGGAGGCCAGCTCAAACATR-ATCAAGAAGAGTAGTACCGCTAGCATTACRpl13a: F-TCTGGAGGACTGTAAGAGGTATGCR-AGACGCACAATCTTGAGAGCAG

RNA sequencing and analysis. RNA for sequencing was isolated from 50 WT orMZudu−/− embryos per sample at tailbud stage according to instructions for theDynabeads mRNA direct kit (Ambion). Embryos from two clutches per genotype(e.g., WT A and B) were collected, then divided in two (e.g., WT A1, A2, B1, andB2) to yield four independently prepared libraries representing two biological andtwo technical replicates per genotype. Libraries for were prepared according toinstructions for the Epicentre ScriptSeq v2 RNA-seq Library preparation kit(Illumina). Briefly, RNA was enzymatically fragmented prior to cDNA synthesis.cDNA was then 3′ tagged, purified using Agencourt AMPure beads, and PCRamplified, at which time sequencing indexes were added. Indexed libraries werethen purified and submitted to the Washington University Genome TechnologyAccess Center for sequencing using an Illumina HiSeq 2500 to obtain single-ended50 bp reads. Raw reads were mapped to the zebrafish GRCz10 reference genomeusing STAR (2.4.2a)61 with default parameters. FeatureCounts (v1.4.6) from theSubread package62 was used to quantify the number of uniquely mapped reads(phred score ≥10) to gene features based on the Ensembl annotations (v83). Sig-nificantly differentially expressed genes were determined by using DESeq2 in thenegative binomial distribution model63 with a cutoff of adjusted p-value ≤0.05 andfold change ≥2.0. Heatmaps were built using the heatmaps2 package in R, andother plots were built using the ggplot2 package in R.

DamID-seq. E. coli DNA adenine methyltransferase (EcoDam) was cloned fromthe pIND(V5) EcoDam plasmid21 (a kind gift from Dr. Bas Van Steensel, Neth-erlands Cancer Institute) by Gibson assembly64 into a 3′ Gateway entry vectorcontaining 6 Myc tags and an SV40 polyA signal (p3E-MTpA). The resulting p3E-EcoDam-MTpA vector was then Gateway cloned into PCS2+ downstream of uducDNA or eGFP to produce C terminal fusions. The resulting udu-dam-myc andgfp-dam-myc plasmids were linearized by KpnI digestion and transcribed using themMessage mMachine SP6 in vitro transcription kit (Ambion). WT AB* embryoswere injected at one-cell stage with 3 pg mRNA encoding Gon4l-Dam-Myc or with1 pg encoding GFP-Dam-Myc as controls. The remainder of the protocol wascarried out largely as in ref. 21 with modifications. Genomic (g)DNA was collectedat tailbud stage using a Qiagen DNeasy kit with the addition of RNAse A. gDNAwas digested with DpnI overnight, followed by ligation of DamID adaptors. Un-methylated regions were destroyed by digestion with DpnII, and then methylatedregions were amplified using primers complementary to DamID adaptors. Twoidentical 50 μl PCR reactions were performed and pooled for each sample to reduceamplification bias, and three independent biological replicates were collected percondition. Amplicons were purified using a Qiagen PCR purification kit, thendigested with DpnII to remove DamID adaptors. Finally, samples were purifiedusing Agencourt AMPure XP beads and submitted for library preparation andsequencing at the Washington University GTAC using an Illumina HiSeq 2500 toobtain single-ended 50 bp reads.

Analysis of DamID-seq data. Raw reads were aligned to zebrafish genomeGRCz10 by using bwa mem (v0.7.12) with default parameters65, then sorted andconverted into bam format by using SAMtools (v1.2)66. The zebrafish genome wasdivided into continuous 1000 bp bins, and FeatureCounts (v1.4.6) from the Sub-read package62 was used to quantify the number of uniquely mapped reads (phredscore ≥10) in each bin. Significantly differentially Gon4l-associated bins weredetermined by using DESeq2 in the negative binomial distribution model63 withstringent cutoff: adjusted p-value ≤0.01 and fold change ≥4.0. Promoter regionswere defined as 2 kb upstream of the transcription start site of a gene based onEnsembl annotations (v83). Wiggle and bigwig files were created from bam filesusing IGVtools (v2.3.60) (https://software.broadinstitute.org/software/igv/igvtools)



https://software.broadinstitute.org/software/igv/igvtools


and wigToBigWig (v4)67, respectively. BigWig tracks were visualized using IGV(v2.3.52)68.

Statistical analysis. Graphpad Prism 6 and 7 software was used to perform sta-tistical analyses and generate graphs of data collected from embryo images. Thestatistical tests used varied as appropriate for each experiment and are described inthe text and figure legends. Data were tested for normal distribution, and non-parametric tests (Mann–Whitney and Kolmogorov–Smirnov) were used for allnon-normally distributed data. Normally distributed data with similar variancebetween groups were analyzed using parametric tests (T-tests and ANOVAs). Alltests used were two tailed. Differential expression and differential enrichmentanalysis of RNA and DamID sequencing data were completed as described above.Panther69 was used to classify differentially expressed genes and to produce piecharts; DAVID Bioinformatic Resources70 was used for functional annotationanalysis.

Subcloning. The full-length udu open reading frame was subcloned from WTcDNA using following primers:

udu cacc F1: CACCATGGGATGGAAACGCAAGTCTTCudu TGA R: TCAGTCCTGCTCTTCATCAGTGGCudu R: GTCCTGCTCTTCATCAGTGGCCGACudu cDNAs with or without a stop codon were cloned into the pENTR/D-

TOPO (Thermo Fisher) vector, which were then Gateway cloned into PCS2+upstream of polyA signal or eGFP to produce a C terminal fusion.

The full-length epcam open reading frame was subcloned from WT cDNAusing the following primers:

epcam F: GGATCCCATCGATTCGATGAAGGTTTTAGTTGCCTTGepcam R: ACTCGAGAGGCCTTGTTAAGAAATTGTCTCCATCTCThe 5′ portion of the itga3b open reading frame was cloned from WT cDNA,

and the 3′ portion was cloned from a partial cDNA clone (GE/Dharmacon) usingthe following primers:

5′ itga3b F: TTTGCAGGCGCGCCGGATCCCATCGATTCGATGGCCGGAAAGTCTCTG

5′ itga3b R: ATTTGAGTGAGTATGGAATGGAGATGTTGAGCG3′ itga3b F: TCAACATCTCCATTCCATACTCACTCAAATACTCAGG3′ itga3b R: GTTCTAGAGGTTTAAACTCGAGAGGCCTTGTCAGAACTC

CTCCGTCAGThe resulting amplicons were Gibson cloned64 into PJS2 (a derivative of

PCS2+) linearized with EcoRI. To create an Itgα3b-GFP fusion, the itga3b openreading frame minus the stop codon was amplified from this plasmid and Gibsoncloned upstream of an eGFP open reading frame into PJS2.

Data availability. The authors declare that all data supporting the findings of thisstudy are available within the article and its Supplementary Information files orfrom the corresponding author upon reasonable request.

Processed RNA-seq and DamID-seq data are available in Supplementary Datafiles 1-3 of this publication, and raw data have been deposited in the GeneExpression Omnibus (GEO) database under the accession codes:

GSE96575 (for RNA-seq) (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=ipwzmaoqjpajbqd&acc=GSE96575) and GSE96576 (for DamID-seq)(https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=apkloooyvxsjxij&acc=GSE96576).

Received: 13 June 2017 Accepted: 6 March 2018

References1. Shih, J. & Keller, R. Patterns of cell motility in the organizer and dorsal

mesoderm of Xenopus laevis. Development 116, 915–930 (1992).2. Warga, R. M. & Kimmel, C. B. Cell movements during epiboly and

gastrulation in zebrafish. Development 108, 569–580 (1990).3. Shih, J. & Keller, R. Cell motility driving mediolateral intercalation in explants

of Xenopus laevis. Development 116, 901–914 (1992).4. Sepich, D. S., Calmelet, C., Kiskowski, M. & Solnica-Krezel, L. Initiation of

convergence and extension movements of lateral mesoderm during zebrafishgastrulation. Dev. Dyn. 234, 279–292 (2005).

5. Keller, R. et al. Mechanisms of convergence and extension by cell intercalation.Philos. Trans. R. Soc. Lond. B Biol. Sci. 355, 897–922 (2000).

6. Wallingford, J. B. et al. Dishevelled controls cell polarity during Xenopusgastrulation. Nature 405, 81–85 (2000).

7. Jessen, J. R. et al. Zebrafish trilobite identifies new roles for Strabismus ingastrulation and neuronal movements. Nat. Cell Biol. 4, 610–615 (2002).

8. Topczewski, J. et al. The zebrafish glypican knypek controls cell polarityduring gastrulation movements of convergent extension. Dev. Cell 1, 251–264(2001).

9. Williams, M., Yen, W., Lu, X. & Sutherland, A. Distinct apical and basolateralmechanisms drive planar cell polarity-dependent convergent extension of themouse neural plate. Dev. Cell 29, 34–46 (2014).

10. Lin, F. et al. Essential roles of Gα12/13 signaling in distinct cell behaviorsdriving zebrafish convergence and extension gastrulation movements. J. CellBiol. 169, 777–787 (2005).

11. Solnica-Krezel, L. et al. Mutations affecting cell fates and cellularrearrangements during gastrulation in zebrafish. Development 123, 67–80(1996).

12. Myers, D. C., Sepich, D. S. & Solnica-Krezel, L. Bmp activity gradient regulatesconvergent extension during zebrafish gastrulation. Dev. Biol. 243, 81–98(2002).

13. Chen, T. & Dent, S. Y. Chromatin modifiers and remodellers: regulators ofcellular differentiation. Nat. Rev. Genet. 15, 93–106 (2014).

14. Ma, Y. et al. The chromatin remodeling protein Bptf promotes posteriorneuroectodermal fate by enhancing Smad2-activated wnt8a expression. J.Neurosci. 35, 8493-8506 (2015).

15. Brownfield, L. et al. Arabidopsis DUO POLLEN3 is a key regulator of malegermline development and embryogenesis. Plant Cell 21, 1940–1956 (2009).

16. Friedman, L., Santa Anna-Arriola, S., Hodgkin, J. & Kimble, J. gon-4, a celllineage regulator required for gonadogenesis in Caenorhabditis elegans. Dev.Biol. 228, 350–362 (2000).

17. Bulchand, S., Menon, S. D., George, S. E. & Chia, W. Muscle wasted: a novelcomponent of the Drosophila histone locus body required for muscle integrity.J. Cell Sci. 123, 2697-2707 (2010).

18. Liu, Y. et al. The zebrafish udu gene encodes a novel nuclear factor and isessential for primitive erythroid cell development. Blood 110, 99–106 (2007).

19. Lu, P. et al. The Justy mutation identifies Gon4-like as a gene that is essentialfor B lymphopoiesis. J. Exp. Med. 207, 1359–1367 (2010).

20. van Steensel, B. & Henikoff, S. Identification of in vivo DNA targets ofchromatin proteins using tethered dam methyltransferase. Nat. Biotechnol. 18,424–428 (2000).

21. Vogel, M. J., Peric-Hupkes, D. & van Steensel, B. Detection of in vivoprotein–DNA interactions using DamID in mammalian cells. Nat. Protoc. 2,1467–1478 (2007).

22. Solnica-Krezel, L., Schier, A. F. & Driever, W. Efficient recovery of ENU-induced mutations from the zebrafish germline. Genetics 136, 1401–1420(1994).

23. Mullins, M. C., Hammerschmidt, M., Haffter, P. & Nüsslein-Volhard, C.Large-scale mutagenesis in the zebrafish: in search of genes controllingdevelopment in a vertebrate. Curr. Biol. 4, 189–202 (1994).

24. Marlow, F. et al. Functional interactions of genes mediating convergentextension, knypek and trilobite, during the partitioning of the eye primordiumin zebrafish. Dev. Biol. 203, 382–399 (1998).

25. Knapik, E. W. et al. A microsatellite genetic linkage map for zebrafish (Daniorerio). Nat. Genet. 18, 338–343 (1998).

26. Hammerschmidt, M. et al. Mutations affecting morphogenesis duringgastrulation and tail formation in the zebrafish, Danio rerio. Development 123,143–151 (1996).

27. Ciruna, B. et al. Production of maternal-zygotic mutant zebrafish by germ-linereplacement. Proc. Natl Acad. Sci. USA 99, 14919-14924 (2002).

28. Lim, C. H., Chong, S. W. & Jiang, Y. J. Udu deficiency activates DNA damagecheckpoint. Mol. Biol. Cell 20, 4183-4193 (2009).

29. Blagden, C. S., Currie, P. D., Ingham, P. W. & Hughes, S. M. Notochordinduction of zebrafish slow muscle mediated by Sonic hedgehog. Genes Dev.11, 2163-2175 (1997).

30. Payne, E. M. et al. Ddx18 is essential for cell-cycle progression in zebrafishhematopoietic cells and is mutated in human AML. Blood 118, 903–915(2011).

31. Liu, Y., Sepich, D. S. & Solnica-Krezel, L. Stat3/Cdc25a-dependent cellproliferation promotes embryonic axis extension during zebrafishgastrulation. PLoS Genet. 13, e1006564 (2017).

32. Park, M. & Moon, R. T. The planar cell-polarity gene stbm regulates cellbehaviour and cell fate in vertebrate embryos. Nat. Cell Biol. 4, 20–25 (2002).

33. Yin, C., Kiskowski, M., Pouille, P. A., Farge, E. & Solnica-Krezel, L.Cooperation of polarized cell intercalations drives convergence and extensionof presomitic mesoderm during zebrafish gastrulation. J. Cell Biol. 180,221–232 (2008).

34. Ciruna, B., Jenny, A., Lee, D., Mlodzik, M. & Schier, A. F. Planar cell polaritysignalling couples cell division and morphogenesis during neurulation. Nature439, 220–224 (2006).

35. Veeman, M. T. et al. Chongmague reveals an essential role for laminin-mediated boundary formation in chordate convergence and extensionmovements. Development 135, 33–41 (2008).

36. Lu, P. et al. The developmental regulator protein Gon4l associates with proteinYY1, co-repressor Sin3a, and histone deacetylase 1 and mediatestranscriptional repression. J. Biol. Chem. 286, 18311-18319 (2011).



https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=ipwzmaoqjpajbqd&acc=GSE96575

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=ipwzmaoqjpajbqd&acc=GSE96575

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=apkloooyvxsjxij&acc=GSE96576

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=apkloooyvxsjxij&acc=GSE96576



37. Maghzal, N., Vogt, E., Reintsch, W., Fraser, J. S. & Fagotto, F. The tumor-associated EpCAM regulates morphogenetic movements through intracellularsignaling. J. Cell Biol. 191, 645–659 (2010).

38. Maghzal, N., Kayali, H. A., Rohani, N., Kajava, A. V. & Fagotto, F. EpCAMcontrols actomyosin contractility and cell adhesion by direct inhibition ofPKC. Dev. Cell 27, 263–277 (2013).

39. Nishiuchi, R. et al. Characterization of the ligand-binding specificities ofintegrin alpha3beta1 and alpha6beta1 using a panel of purified lamininisoforms containing distinct alpha chains. J. Biochem. 134, 497–504 (2003).

40. Slanchev, K. et al. The epithelial cell adhesion molecule EpCAM is requiredfor epithelial morphogenesis and integrity during zebrafish epiboly and skindevelopment. PLoS Genet. 5, e1000563 (2009).

41. Carney, T. J. et al. Genetic analysis of fin development in zebrafish identifiesfurin and hemicentin1 as potential novel fraser syndrome disease genes. PLoSGenet. 6, e1000907 (2010).

42. Webb, A. E. et al. Laminin alpha5 is essential for the formation of thezebrafish fins. Dev. Biol. 311, 369–382 (2007).

43. Thisse, B. & Thisse, C. ZFIN Direct Data Submission http://zfin.org (2004).44. Pollard, S. M. et al. Essential and overlapping roles for laminin alpha chains in

notochord and blood vessel formation. Dev. Biol. 289, 64–76 (2006).45. Ramos, J. W., Whittaker, C. A. & DeSimone, D. W. Integrin-dependent

adhesive activity is spatially controlled by inductive signals at gastrulation.Development 122, 2873-2883 (1996).

46. Fagotto, F., Rohani, N., Touret, A. S. & Li, R. A molecular base for cell sortingat embryonic boundaries: contact inhibition of cadherin adhesion by ephrin/Eph-dependent contractility. Dev. Cell 27, 72–87 (2013).

47. Rauzi, M., Verant, P., Lecuit, T. & Lenne, P. F. Nature and anisotropy ofcortical forces orienting Drosophila tissue morphogenesis. Nat. Cell Biol. 10,1401–1410 (2008).

48. Shindo, A. & Wallingford, J. B. PCP and septins compartmentalize corticalactomyosin to direct collective cell movement. Science 343, 649–652 (2014).

49. Suzuki, A. & Itoh, T. Effects of calyculin A on tension and myosinphosphorylation in skinned smooth muscle of the rabbit mesenteric artery. Br.J. Pharmacol. 109, 703–712 (1993).

50. Kilian, B. et al. The role of Ppt/Wnt5 in regulating cell shape and movementduring zebrafish gastrulation. Mech. Dev. 120, 467–476 (2003).

51. Nikaido, M., Law, E. W. & Kelsh, R. N. A systematic survey of expression andfunction of zebrafish frizzled genes. PLoS ONE 8, e54833 (2013).

52. Green, J. B., Dominguez, I. & Davidson, L. A. Self-organization of vertebratemesoderm based on simple boundary conditions. Dev. Dyn. 231, 576–581(2004).

53. Monier, B., Pélissier-Monier, A., Brand, A. H. & Sanson, B. An actomyosin-based barrier inhibits cell mixing at compartmental boundaries in Drosophilaembryos. Nat. Cell Biol. 12, 60–65 (2010). sup pp. 61–69.

54. Umetsu, D. et al. Local increases in mechanical tension shape compartmentboundaries by biasing cell intercalations. Curr. Biol. 24, 1798–1805 (2014).

55. Westerfield, M. The Zebrafish Book. A Guide for the Laboratory Use ofZebrafish (Danio rerio) (University of Oregon Press, Eugene, OR, 1993).

56. Kimmel, C. B., Ballard, W. W., Kimmel, S. R., Ullmann, B. & Schilling, T. F.Stages of embryonic development of the zebrafish. Dev. Dyn. 203, 253–310(1995).

57. Köprunner, M., Thisse, C., Thisse, B. & Raz, E. A zebrafish nanos-related geneis essential for the development of primordial germ cells. Genes Dev. 15, 2877-2885 (2001).

58. Neff, M. M., Neff, J. D., Chory, J. & Pepper, A. E. dCAPS, a simpletechnique for the genetic analysis of single nucleotide polymorphisms:experimental applications in Arabidopsis thaliana genetics. Plant J. 14,387–392 (1998).

59. Thisse, C. & Thisse, B. High-resolution in situ hybridization to whole-mountzebrafish embryos. Nat. Protoc. 3, 59–69 (2008).

60. Aigouy, B., Umetsu, D. & Eaton, S. Segmentation and quantitative analysis ofepithelial tissues. Methods Mol. Biol. 1478, 227–239 (2016).

61. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29,15–21 (2013).

62. Liao, Y., Smyth, G. K. & Shi, W. The Subread aligner: fast, accurateand scalable read mapping by seed-and-vote. Nucleic Acids Res. 41, e108(2013).

63. Love, M. I., Huber, W. & Anders, S. Moderated estimation of foldchange and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550(2014).

64. Gibson, D. G. et al. Enzymatic assembly of DNA molecules up to severalhundred kilobases. Nat. Methods 6, 343–345 (2009).

65. Li, H. & Durbin, R. Fast and accurate short read alignment withBurrows–Wheeler transform. Bioinformatics 25, 1754–1760 (2009).

66. Li, H. et al. The Sequence Alignment/Map format and SAMtools.Bioinformatics 25, 2078–2079 (2009).

67. Kent, W. J., Zweig, A. S., Barber, G., Hinrichs, A. S. & Karolchik, D. BigWigand BigBed: enabling browsing of large distributed datasets. Bioinformatics 26,2204-2207 (2010).

68. Thorvaldsdóttir, H., Robinson, J. T. & Mesirov, J. P. Integrative GenomicsViewer (IGV): high-performance genomics data visualization and exploration.Brief. Bioinform 14, 178–192 (2013).

69. Mi, H., Muruganujan, A., Casagrande, J. T. & Thomas, P. D. Large-scale genefunction analysis with the PANTHER classification system. Nat. Protoc. 8,1551–1566 (2013).

70. Huang, da W., Sherman, B. T. & Lempicki, R. A. Systematic and integrativeanalysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc.4, 44–57 (2009).

AcknowledgementsWe thank Dr. Bas Van Steensel for EcoDam plasmids, Drs. Christine and Bernard Thissefor WISH probes, Dr. Bo Zhang and Dr. Scott Higdon for bioinformatics help,Dr. Matthew Hass for DamID advice, Bisiayo Fashemi for assistance with image analysis,the Washington University Center for Cellular Imaging for assistance with laser ablationexperiments, and the Washington University Genome Technology Access Center forlibrary preparation and sequencing services. The National Institutes of Health grantsR01GM55101 and R35GM118179 to L.S.-K. and F32GM113396 to M.L.K.W., and aW.M. Keck Foundation Fellowship to M.L.K.W. in part supported this study.

Author contributionsM.L.K.W., A.S., and L.S.-K. designed the study. M.L.K.W., A.S., and T.B. performedexperiments. C.Y. participated in the initial forward genetic screen. P.G. performedbioinformatic analysis. M.L.K.W. and L.S.-K. wrote the manuscript.

Additional informationSupplementary Information accompanies this paper at https://doi.org/10.1038/s41467-018-03715-w.

Competing interests: The authors declare no competing interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,

adaptation, distribution and reproduction in any medium or format, as long as you giveappropriate credit to the original author(s) and the source, provide a link to the CreativeCommons license, and indicate if changes were made. The images or other third partymaterial in this article are included in the article’s Creative Commons license, unlessindicated otherwise in a credit line to the material. If material is not included in thearticle’s Creative Commons license and your intended use is not permitted by statutoryregulation or exceeds the permitted use, you will need to obtain permission directly fromthe copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

© The Author(s) 2018



http://zfin.org

https://doi.org/10.1038/s41467-018-03715-w

https://doi.org/10.1038/s41467-018-03715-w

http://npg.nature.com/reprintsandpermissions/

http://npg.nature.com/reprintsandpermissions/

http://creativecommons.org/licenses/by/4.0/

http://creativecommons.org/licenses/by/4.0/


Gon4l regulates notochord boundary formation and cell ...€¦ · notochord boundary via...

Documents

Transcript of Gon4l regulates notochord boundary formation and cell ...€¦ · notochord boundary via...